AI Conversational Interviews

Write your questions once, get one share link, and send it to customers. An AI model conducts the conversation turn by turn — it asks a question, reads the answer, and follows up on what the person actually said instead of marching through a static form. You get back a transcript, an AI summary, and a structured answer for every question you planned.

Interviews live at /interviews in the dashboard. They are team-scoped: every member of the team sees the same interviews and the same responses.

---

How it works

1. You write the questions. A plain list of the things you want to learn — no branching logic, no skip rules to configure.
2. Zenovay gives you one share link. That link is the whole distribution mechanism. Put it in an email, a Slack message, a support reply, or a post-purchase note.
3. The AI conducts the conversation. The respondent answers in a chat window. The AI works through your planned questions but reacts to each answer — it asks "why?", clarifies a vague reply, and drops a follow-up that no static form could have known to ask.
4. The response is stored as a transcript. Not a row of {question, answer} pairs — the full back-and-forth, so you can read what was actually said and in what order.
5. The AI extracts the structure afterwards. For every question you planned, Zenovay produces an answer, a confidence indicator, and the verbatim quote from the transcript that supports it. If the respondent never really answered question 4, the confidence says so.
6. You chat with the results. Ask a follow-up question of a single response, or across all responses at once.

---

Create an interview

1. Open Interviews in the dashboard sidebar.
2. Click New interview and give it a name — respondents see this name, so write it for them, not for you.
3. Add your questions, one per line. Keep them open-ended; the AI handles the drilling-down, so "What made you cancel?" works better than a multiple-choice list.
4. Optionally write the intro the respondent sees when they open the link — a sentence explaining who you are and how long this will take meaningfully improves completion.
5. Save, then copy the share link and send it out.

The interview counts against your plan's active-interview limit while it is live. Archive an interview you are finished with to free the slot; archiving keeps the responses.

---

Reading the results

Each response has three layers, from fastest to most detailed:

The confidence indicator matters more than it looks. A conversational interview does not guarantee that every planned question gets a clean answer — people go off on tangents, and that is often the valuable part. Low confidence on a question is a signal to open the transcript, not a bug.

Ask about the results

Two entry points, both powered by the same chat:

• Ask about this response — scoped to one respondent. Good for "did they say anything about pricing?"
• Ask about all responses — scoped across every response to the interview. Good for "what is the most common reason people gave for churning?"

Both are rate-limited per team per day; the quota is in the plan table below.

---

Voice mode

Pro and above

On Pro, Scale and Enterprise you can enable voice mode, which turns the share link into a spoken interview: the respondent talks, the AI listens and replies out loud, and the conversation is transcribed as it goes. Everything downstream is identical — the same transcript storage, the same summary, the same per-question extraction.

Voice raises completion quality on open-ended research (people say more out loud than they type) and lowers it on anything requiring precision, because a spoken answer is transcribed and transcription is imperfect. Use it for discovery, not for collecting order numbers.

Voice mode is off by default. Enable it per interview when you create or edit it.

---

Limits per plan

Two more gates worth knowing:

• Completion webhooks and the confirmation email to the respondent start on Pro.
• Password-protecting the share link starts on Scale.

Archived interviews do not count toward the active-interview limit.

---

What this is not

Interviews are a standalone share-link product. They are not an on-site widget, and this distinction is worth being explicit about:

• Zenovay does not offer in-page surveys, on-site pop-up questionnaires, NPS widgets, CSAT widgets or feedback widgets. There is no snippet that renders a question box on your website.
• An interview is not embedded in your product. The respondent opens the link and has a conversation in a Zenovay-hosted page.
• Interviews are not a replacement for a micro-survey. Asking a visitor a single question at the moment of hesitation and asking a customer twelve minutes of open questions are different jobs, and Interviews only does the second one.

What Interviews replaces is the customer-research call and the long-form feedback form: the conversation you would otherwise have to schedule, run and write up yourself.

---

Practical notes

• Send the link to people who already know you. Response rates on a cold conversational interview are poor; on a link sent to existing customers after a support interaction or a cancellation, they are good.
• Fewer questions, better answers. Five open questions the AI can dig into beats twenty closed ones. The Free limit of 5 is not as restrictive in practice as it sounds.
• Read one full transcript before you trust the summaries. The extraction is good, but you calibrate your trust in it by reading the raw conversation at least once per interview.
• Archive rather than delete. Deleting an interview removes its responses with it.

---

Next Steps

• Session Replay - Watch what people actually did, then ask them why
• Conversion Funnels - Find the step people drop at before you interview them about it
• Goals - Measure whether what you changed after the interviews worked